Focusing on the travel problem of blind people, an audio-visual fusion blind guiding robot system is designed, which can realize the detection of environmental obstacles, the identification of road traffic signs, the planning of path and the real-time interaction of information. In terms of hardware, the real-time visual information are collected through various types of optical sensors, and the information interaction with the users are carried out in the voice form. In terms of software, the neural network algorithm is used to carry out the data fusion from multi-sensors so as to realize the environmental information recognition, and the double pole coefficient method and the image segmentation based on the color histogram are used to realize the recognition of the zebra crossing and the blind road, respectively. The path planning is implemented based on the artificial potential field method. To verify the effectiveness of the proposed scheme, we build a prototype system and test it in a real environment. The test results show that the blind guiding robot has the advantages of high recognition accuracy and robust performance to complex environment, and it is satisfying the travel needs of blind people.